26 Commits

Author SHA1 Message Date
Alexey Samsonov
2429a8f71f
[libc] Add missing and correct some existing C23 functions to math.h (#156512)
This change fixes and closes some gaps in the YAML template for
producing the math.h header.

It adds some missing declarations (dadd/dsub function variants), correct
arguments and/or return type for other functions from this family (dsqrt
and ddiv), and add a missing fminimum_numl variant.
2025-09-02 13:42:48 -07:00
Mohamed Emad
c1d1e0e32f
[libc][math][c23] Implement C23 math function atanpif16 (#150400)
This PR implements `atanpif16(x)` which computes
$\frac{\arctan(x)}{\pi}$ for half-precision floating-point numbers using
polynomial approximation with domain reduction.

## Mathematical Implementation

The implementation uses a 15th-degree Taylor polynomial expansion of
$\frac{\arctan(x)}{\pi}$ that's computed using
[`python-sympy`](https://www.sympy.org/en/index.html) and it's accurate
in $|x| \in [0, 0.5)$:

$$
g(x) = \frac{\arctan(x)}{\pi} \approx 
\begin{aligned}[t]
    & 0.318309886183791x \\
    & - 0.106103295394597x^3 \\
    & + 0.0636619772367581x^5 \\
    & - 0.0454728408833987x^7 \\
    & + 0.0353677651315323x^9 \\
    & - 0.0289372623803446x^{11} \\
    & + 0.0244853758602916x^{13} \\
    & - 0.0212206590789194x^{15} + O(x^{17})
\end{aligned}
$$


--- 

To ensure accuracy across all real inputs, the domain is divided into
three cases with appropriate transformations:

**Case 1: $|x| \leq 0.5$**  
Direct polynomial evaluation: 

$$\text{atanpi}(x) = \text{sign}(x) \cdot g(|x|)$$

**Case 2: $0.5 < |x| \leq 1$**  
Double-angle reduction using:

$$\arctan(x) = 2\arctan\left(\frac{x}{1 + \sqrt{1 + x^2}}\right)$$

$$\text{atanpi}(x) = \text{sign}(x) \cdot 2g\left(\frac{|x|}{1 + \sqrt{1
+ x^2}}\right)$$

**Case 3: $|x| > 1$**  
Reciprocal transformation using 

$$\arctan(x) = \frac{\pi}{2} - \arctan\left(\frac{1}{x}\right) \
\text{for} \ x \gt 0$$

$$\text{atanpi}(x) = \text{sign}(x) \cdot \left(\frac{1}{2} -
g\left(\frac{1}{|x|}\right)\right)$$


Closes #132212
2025-09-02 22:50:50 +03:00
Leandro Lacerda
9dc66a631e
[libc] Add half-precision math functions to the math header (#155060)
This patch adds the `exp10f16`, `exp2f16`, `expf16`, and `expm1f16` math
functions to the `math.h` header.
2025-08-22 22:06:12 -05:00
Leandro Lacerda
15a192cde5
[libc] Enable double math functions on the GPU (#154857)
This patch adds the `acos` math function to the NVPTX build. It also
adds the `sincos` math function to the `math.h` header.
2025-08-22 06:52:13 -05:00
Mohamed Emad
40833eea21
Reland "[libc][math][c23] Implement C23 math function asinpif16" (#152690)
#146226 with fixing asinpi MPFR number function and make it work when
mpfr < `4.2.0`
2025-08-18 00:04:47 +03:00
Leandro Lacerda
cd19fbad09
[libc] Enable float math functions on the GPU (#151841)
This patch adds the `tanpif` math functions to the GPU build. It also
adds the `cospif` and `sinpif` math functions to the `math.h` header.
2025-08-02 22:13:04 -05:00
OverMighty
6a85f7cef3
Revert "[libc][math][c23] Implement C23 math function asinpif16" (#150756)
Reverts llvm/llvm-project#146226

The MPFR test uses `mpfr_asinpi` which requires MPFR 4.2.0 or later, but
the Buildbots are running an older version of MPFR.

See https://lab.llvm.org/buildbot/#/builders/104/builds/27743 for
example.

I said I was going to revert the PR until we have a workaround for older
versions of MPFR, but then I forgot and I just disabled the entrypoints
which doesn't fix the Buildbot builds.
2025-07-26 15:39:43 +03:00
Mohamed Emad
eed9b4e058
[libc][math][c23] Implement C23 math function asinpif16 (#146226)
The function is implemented using the following Taylor series that's
generated using [python-sympy](https://www.sympy.org/en/index.html), and
it is very accurate for |x| $$\in [0, 0.5]$$ and has been verified using
Geogebra. The range reduction is used for the rest range (0.5, 1].

$$
\frac{\arcsin(x)}{\pi} \approx 
\begin{aligned}[t]
    &  0.318309886183791x  \\
    & + 0.0530516476972984x^3 \\
    & + 0.0238732414637843x^5 \\
    & + 0.0142102627760621x^7 \\
    & + 0.00967087327815336x^9 \\
    & + 0.00712127941391293x^{11} \\
    & + 0.00552355646848375x^{13} \\
    & + 0.00444514782463692x^{15} \\
    & + 0.00367705242846804x^{17} \\
    & + 0.00310721681820837x^{19} + O(x^{21})
\end{aligned}
$$

## Geogebra graph

![28-06-2025-1913-eDP-1](https://github.com/user-attachments/assets/f70818e1-1b34-406e-962a-a30fdc909f18)

Closes #132210
2025-07-26 14:02:18 +03:00
Muhammad Bassiouni
c59e4b5805
[libc][math] fix header generation (#149918) 2025-07-22 01:13:14 +03:00
Connector Switch
c3abe3ff22
[libc][math][c23] implement C23 math function tanpif (#147192)
The smoke test and exhaustive test pass on x86_64 Linux.

Closes #94895.
2025-07-12 12:39:33 +08:00
William Huynh
fd432151a6
[libc] Fix bugs found when testing with all headers (#144049)
Fixes a couple of bugs found when building. The PR to enable the headers
can be found here: #144114.

- math.yaml: float128 guard
- wchar.yaml: __restrict keyword order
2025-06-13 10:26:40 -07:00
wldfngrs
573545c712
[libc][math][c23] Add atanf16() function (#141612)
- Implementation of atan (tan inverse) function for 16-bit inputs.
- Exhaustive tests across the 16-bit input range
2025-06-01 13:36:16 +02:00
lntue
78cc822aa6
[libc][math] Implement double precision acos correctly rounded for all rounding modes. (#138308)
We reduce computation of `acos` to `asin` as follow:

When `|x| < 0.5`:
```math
acos(x) = \frac{\pi}{2} - asin(x).
```
For `0.5 <= |x| < 1`, let
```math
u = \frac{1 - \left| x \right|}{2},
```
then
```math
acos(x) = \begin{cases}
  2 \cdot asin \left( \sqrt{u} \right) &, 0.5 \leq x < 1 \\
  \pi - 2 \cdot asin \left( \sqrt{u} \right) &, -1 < x \leq 0.5 
\end{cases}
```
2025-05-08 23:23:09 -04:00
Harrison Hao
accee2b553
[libc][math][c23] Add atanhf16 C23 math function. (#132612)
Implementation of atanhf16 function for 16-bit inputs.

Closes: https://github.com/llvm/llvm-project/issues/132209
2025-04-25 07:53:52 -04:00
Anton
851f7c7421
[libc][math][c23] Add acospif16() function (#134664)
Addresses #132211  #132754
Part of #95250
2025-04-24 18:03:24 -04:00
lntue
8741412bdf
[libc][math] Implement a fast pass for atan2f128 with 1ULP error using DyadicFloat<128>. (#133150)
Part of https://github.com/llvm/llvm-project/issues/131642.
2025-04-01 10:57:32 -04:00
Tejas Vipin
8078665bca
[libc][math][c23] Add hypotf16 function (#131991)
Implement hypot for Float16 along with tests.
2025-03-31 10:06:28 -04:00
Tejas Vipin
d22e35be17
[libc][math][c23] Add asinhf16 function (#131351)
Implement asinh for Float16 along with tests. Closes #131001
2025-03-29 13:54:52 +01:00
Harrison Hao
445837a363
[libc][math][c23] Add fmaf16 C23 math function. (#130757)
Implementation of fmaf16 function for 16-bit inputs.
2025-03-23 10:48:56 +08:00
Harrison Hao
ee3e17d67f
[libc][math][c23] Add acoshf16 C23 math function. (#130588)
Implementation of acoshf16 function for 16-bit inputs.
2025-03-22 22:08:34 -04:00
lntue
a17b03f0e4
[libc][math] Implement fast pass for double precision atan function. (#132333)
Implement fast pass for double precision `atan` using range reduction
modulo 1/64 and degree-9 Taylor polynomial.
Relative error is bounded by 2^-66.
2025-03-21 14:12:06 -04:00
wldfngrs
3e284554a8
[libc][math][c23] Add acosf16() function (#127731)
- Implementation of acosf16 (inverse cosine) function for 16-bit inputs.
- Exhaustive tests across the 16-bit input range.
2025-02-23 19:46:46 +01:00
wldfngrs
7ee56b9afc
[libc][math][c23] Add asinf16() function (#124212)
Co-authored-by: OverMighty <its.overmighty@gmail.com>
2025-02-10 12:38:55 +01:00
wldfngrs
ecf4f95c4f
[libc][math][c23] Add tanf16 function (#121018)
- Implementation of tan for 16-bit floating point inputs.
- Exhaustive tests across the 16-bit input range
2025-01-12 23:46:53 -05:00
Roland McGrath
6ad0dcf67f
[libc] Revamp hdrgen command line and build integration (#121522)
This adds a new main command-line entry point for hdrgen, in the
new main.py.  This new interface is used for generating a header.
The old ways of invoking yaml_to_classes.py for other purposes
are left there for now, but `--e` is renamed to `--entry-point`
for consistency with the new CLI.

The YAML schema is expanded with the `header_template` key where
the corresponding `.h.def` file's path is given relative to where
the YAML file is found.  The build integration no longer gives
the `.h.def` path on the command line.  Instead, the script now
emits a depfile that's used by the cmake rules to track that.
The output file is always explicit in the script command line
rather than sometimes being derived from a directory path.
2025-01-07 15:56:12 -08:00
Roland McGrath
62cd050b63
[libc] Move hdrgen yaml files into include/ (#121443)
The .yaml files should live next to the corresponding .h.def
files in libc/include/, rather than next to the implementation of
the tool in libc/utils/hdrgen/.  As with the .h.def files, there
is no need for a yaml/ subdirectory under include/.  This simpler
layout is more natural for maintenance and also simplifies build
integration outside the LLVM CMake build.
2025-01-01 17:03:35 -08:00